Does Affective Disposition Moderate the Relationship Between Job Satisfaction and \bluntary Turnover?

Transcription

1 Journal of Applied Psychology Vol. 78, No. 3, Copyright 1993 by the American Psychological Association, Inc. 002I-9010/93/S3.00 Does Affective Disposition Moderate the Relationship Between Job Satisfaction and \bluntary Turnover? Timothy A. Judge J. Weitz (1952) argued that job dissatisfaction would be more predictive of turnover if it was considered in light of an individual's predisposition to be satisfied with everyday life events. In the present study it was hypothesized that affective disposition moderates the relationship between job satisfaction and voluntary turnover. With data collected from a sample of nurses, support was indicated for the hypothesis. The more positive the disposition of the individual, the stronger the relationship that was observed between job dissatisfaction and turnover. Furthermore, individuals dissatisfied with their jobs but positively disposed to life in general were the individuals most likely to quit. Implications of the results for future research and practice are discussed. Empirical work over the years has clearly established the significant role of job satisfaction in predicting turnover. Carsten and Spector (1987), in a meta-analysis of 47 studies, estimated a corrected correlation between job satisfaction and turnover of-.26 (the 95% confidence interval did not include 0). Because sampling error accounted for only 21 % of the variance in correlations across studies, the authors concluded that room for significant moderator effects existed. In fact, Carsten and Spector found that alternative employment opportunities moderated the relationship between job satisfaction and turnover. On the basis of Carsten and Specter's findings, the potential exists that other variables interact with job satisfaction in predicting turnover. Given the importance of turnover decisions to individuals and organizations (Dalton & Todor, 1979; Mobley, 1982; Staw, 1980), it is surprising that more research concerning potential moderators of the job satisfaction-turnover relationship has not been conducted. A potential, but previously unstudied, moderator of the job satisfaction-turnover relationship was hypothesized by Weitz (1952). He argued that a worker's level of dissatisfaction might be more meaningful if placed in the context of a worker's predisposition to be satisfied in general. Rather than inferring that a certain level of dissatisfaction will induce turnover among all workers, Weitz argued that attempts to improve the prediction of turnover might benefit from considering the individual's disposition. He speculated that if two workers report the same level of job dissatisfaction, the one most likely to quit is the one with the highest predisposition to be happy or satisfied in general. To measure this predisposition, Weitz proposed a "gripe index" that assessed satisfaction with 44 items prevalent in ev- I thank Charles Hulin for his contributions to this project and Bob Bretz, Barry Gerhart, and three anonymous reviewers for comments on an earlier version of this article. Correspondence concerning this article should be addressed to Timothy A. Judge, Department of Personnel and Human Resource Studies, Center for Advanced Human Resource Studies, New York State School 9f Industrial and Labor Relations, Cornell University, 393 Ives Hall, Ithaca, New York eryday life. These items ranged from the way people drive, to income tax, to the national political situation. In essence, Weitz (1952) was hypothesizing an interaction between affective disposition and job satisfaction in predicting turnover. He suggested that "some individuals generally gripe more than others" (p. 203) and that such individuals, when dissatisfied with their jobs, are less likely to quit than are those more positively disposed toward life. Weitz explained that this should be expected because an individual with a positive disposition reporting a certain level of job dissatisfaction is more dissatisfied on the job relative to other things in his or her life than an individual with a negative disposition reporting the same level of job dissatisfaction. This also suggests that the relationship between job dissatisfaction and turnover is greater for generally satisfied than for generally dissatisfied individuals, because for the latter (those with a negative disposition) dissatisfaction with the job is no more meaningful or exceptional than the other dissatisfying events in their lives. Several conceptual and theoretical factors may explain why this hypothesized relationship is reasonable. Mobley's (1977) psychological process model of turnover provides some support for Weitz's (1952) hypothesis. Mobley argued that job dissatisfaction is translated into thoughts of quitting, evaluation of alternatives, and ultimately, turnover because quitting is expected to result in a more satisfying job. However, those more negatively disposed toward life may have no such expectation. For them, job dissatisfaction simply may be another dissatisfying element in an already dissatisfying world. This is supported by research suggesting that job satisfaction may derive from genetic or early childhood influences (Arvey, Bouchard, Segal, & Abraham, 1989; Staw, Bell, & Clausen, 1986). Thus, for these individuals, job dissatisfaction and quitting may seem to have little to do with each other. Changing jobs may not result in higher satisfaction because the dissatisfaction is due less to the characteristics of the job than to affective predispositions. On the other hand, job dissatisfaction is much more salient and generates more tension for generally happy individuals, and changing jobs may appear to be a viable means of correcting one of the few dissatisfying elements in their lives.

2 396 TIMOTHY A. JUDGE A related possibility, and probably one closer to Weitz's (1952) reasoning, is that assessing affective disposition permits a more accurate assessment of the true job dissatisfaction of the individual relative to other things in his or her life. Individuals may act on dissatisfaction with a specific object only when it surpasses some relative internal standard. Generally dissatisfied individuals who report dissatisfaction with their jobs are reporting an average internal level of satisfaction. Conversely, generally satisfied individuals reporting the same level of job dissatisfaction are reporting a much higher degree ofdissatisfaction with their jobs relative to other concerns in their lives. Thus, by accounting for the satisfaction predisposition of individuals, it is possible to obtain a more accurate measurement of the relative degree of job satisfaction or dissatisfaction. As a result, higher relations with turnover are expected to be observed. A third possibility supporting the moderating effect of affective disposition on the job satisfaction-turnover relationship derives from past dispositional research. Staw and Ross (1985) and Staw et al. (1986) have suggested that changes in job conditions may be constrained by dispositional forces. As Gerhart (1987,1990b) pointed out, this implies a dispositional interaction. This dispositional interaction may indicate that the potentially adaptive nature of turnover is less for those with a negative disposition than for those with a positive disposition. Because those with a negative disposition are generally dissatisfied with the world around them, changing the conditions of the job may do little to change this generalized state. As Staw and Ross (1985) pointed out, disposition may affect how individuals rationalize difficult or dissatisfying job conditions. Thus, those with a negative disposition may fail to take action as a result of job dissatisfaction through beliefs that one cannot improve one's life for the better, that a job is simply a means to earn a living and nothing more, and other such cynical rationalizations (Kanter & Mirvis, 1989). A final supporting prediction regarding the interaction between affective disposition and job satisfaction in predicting turnover derives from the possibility of a response set. It is possible that measures of affective disposition assess the propensity to use a response set in completing a job satisfaction questionnaire (e.g., a dispositional measure may simply reflect the tendency to respond positively or negatively when reacting to survey questions). Once the response set is taken into account, a more valid measure of the job satisfaction construct is obtained (i.e., true levels of job satisfaction are measured, taking into account the artifact of individuals' response tendencies), and hence the ability of job satisfaction to predict turnover is enhanced (see Gerhart, 1990b). If this interpretation is correct, the practical implication is that before attitudes are related to such behaviors as turnover, attitude measures should be adjusted for response tendencies (Gerhart, 1990b). Thus, I hypothesize that there is an interaction between affective disposition and job satisfaction in predicting turnover. Specifically, I expect that the effect of job satisfaction on turnover depends on the affective disposition of the individual and that the more positive the disposition of an individual, the stronger the negative relationship between job satisfaction and turnover. In addition to the hypothesized interaction and the necessary main effects of job satisfaction and affective disposition on turnover, I took several relevant control variables into account. I based the selection of these variables on Mobley's (1982) review of past research, which suggested a number of potential influences on turnover. These variables were age (Porter & Steers, 1973), experience (Mobley, 1982), wage rates (Dalton & Todor, 1979), education (Mellow, 1980), and labor market alternatives (Gerhart, 1990a). Setting and Subjects Method The setting for this research was a medical clinic located in the Midwest. Subjects (N = 234) were registered nurses (56%), licensed practical nurses (15%), medical office assistants (15%), and laboratory technicians or clinical specialists (14%). Education of the respondents ranged from a high school diploma (12%) to a master's degree (9%). The average hourly wage rate was $9.57, with a range from $4.15 to $17.43 (SD = $2.69). Age ranged from 21 to 70 years, with an average age of 37.3 years (SD =9.3 years). Job tenure ranged from newly employed to 50 years; the average level of tenure was 12.1 years (SD = 8.0 years). Approximately 40% of respondents perceived little or no employment alternatives, about 42% perceived some alternatives, and about 18% perceived many employment alternatives. The annualized average voluntary turnover rate for subjects in this sample was 17.9%, which is close to the average turnover rate of the U.S. work force (Mobley, 1982). Measures Affective disposition. I measured affective disposition by a scale derived from the measure developed by Weitz (1952). His 44-item scale consisted of a checklist of facially neutral objects. Weitz termed the scale a gripe index because individuals expressing a large number of dissatisfactions in their life, as measured by the checklist, may be predisposed to view most things negatively. Thus, the survey measures disposition by reflecting affective bias toward items common to everyday life. Individuals highly satisfied with the objects as a whole may have a Pollyannaish tendency to see everything (including their job) in a favorable light. The obverse also is hypothesized to be true. The use of Weitz's (1952) measure is an intended departure from research that uses measurements of positive and negative atfectivity (cf. Watson, Clark, & Tellegen, 1988). Although some research supports the distinction between positive and negative affectivity (Brief, Burke, George, Robinson, & Webster, 1988; Watson, 1988; Watson et al, 1988; Watson & Tellegen, 1985), other research suggests that measures of affectivity are not stable over time (as true dispositional measures arguably should be) and that measurements of positive and negative affectivity actually reflect affect experienced (e.g., respondents are asked how often they have experienced states of happiness recently) rather than the disposition toward affect (Diener, 1984, 1990; Judge, 1992). Weitz's (1952) measure was preferred because it avoids the controversial distinction between positive and negative affectivity. It is assumed that Weitz's measure reflects a better assessment of a dispositional trait as opposed to an affective state than do measurements of positive and negative affectivity. In fact, Judge and Bretz (in press) demonstrated that Weitz's measure possessed favorable psychometric properties and displayed greater stability over time than Watson et al.'s (1988) measurement of positive and negative affectivity. Thus, evidence suggests that Weitz's measure is a valid measure of affective disposition and that it should be construed as an assessment of the trait of affective disposition rather than of an affective state. In this study I modified Weitz's (1952) checklist in several ways. The modified, 25-item survey eliminates items that were thought to reflect

3 DISPOSITION, JOB SATISFACTION, AND TURNOVER 397 contaminated measures of disposition. For example, the item "area of city in which you live" was eliminated because it may be confounded with socioeconomic status (e.g., the social and economic background of the individual may be related to the area in which he or she lives). "Your last job" was eliminated because past research has suggested some stability in job characteristics (Gerhart, 1987), thus potentially confounding disposition with previous and present job quality. Items not applicable to all individuals, such as "the college you attended," also were eliminated. In addition, wording was modernized (e.g., automobile was changed to car). Finally, the original dichotomous checklist was changed to a trichotomous response scale (1 = dissatisfied; 2 = neutral; 3 = satisfied). The coefficient alpha reliability estimate for the revised scale was.78. Items from the revised scale, the Neutral Objects Satisfaction Questionnaire, are provided in Table 1. Job satisfaction. Job satisfaction was measured by the Job Descriptive Index (JDI; Smith, Kendall, & Hulin, 1969), as modified by Roznowski (1989). Overall job satisfaction can be represented by five facets: pay, promotion opportunities, supervision, co-workers, and the work itself. The intercorrelations of those facets reveal a communality among the dimensions, suggesting a second-order general factor (Parsons & Hulin, 1982). In the present study, the reliabilities of the JDI subscales ranged from.85 to.91. Overall job satisfaction was constructed by multiplying the pay and promotion scales by 2 and then adding all five subscales. Results from a second-order factor analysis confirmed the ability of the five subscales to represent overall job satisfaction. Voluntary turnover. Data on voluntary turnover were gathered from company records 10 months after surveys were completed by the respondents. Company representatives were asked to indicate which separations were voluntary and which were not. Of the 49 separations that occurred over the 10 months, 36 were voluntary and 13 were dismissals. Other variables. Alternative employment opportunities were assessed by asking each individual to estimate her or his alternative em- Table 1 Items in the Neutral Objects Satisfaction Questionnaire Item No. Item 1. The city in which you live 2. The residence where you live 3. The neighbors you have 4. The high school you attended 5. The climate where you live 6. The movies being produced today 7. The quality of food you buy 8. Today's cars 9. Local newpapers 10. Your relaxation time 11. Your first name 12. The people you know 13. Television programs 14. Local speed limits 15. The way people drive 16. Advertising 17. The way you were raised 18. Telephone service 19. Public transportation 20. Restaurant food 21. Yourself 22. Modern art 23. Popular music 24. 8^ in. X 11 in. paper 25. Your telephone number ployment opportunities at the present time (1 = no alternatives; 5 = many alternatives). Age, wage rate, job tenure, and highest educational level achieved also were measured from specific questions on the survey. Procedure The questionnaire data collected for this study were part of a larger study that served as the basis for my dissertation (Judge, 1990). Surveys were administered to employees on a voluntary basis during their work hours. Department supervisors coordinated the scheduling. Employees gathered in small group sessions (15-20 individuals), where it was explained that the purpose of the study was to understand how workers view their lives and their jobs (to avoid priming effects, it was emphasized in the survey's cover letter that there were no right or wrong answers and that employees should respond as openly and honestly as possible). Employees then were asked (but not required) to participate. Confidentiality of individual responses was assured, and feedback on the results of the survey was promised and subsequently delivered. Three hundred twenty employees worked in the departments that participated in the study; 255 employees completed usable surveys, representing a response rate of 80%. In follow-up conversations, department supervisors suggested that most of those not participating were either on leave or unavailable for other reasons beyond those employees' control. Turnover data were collected after completion of my dissertation. Because of listwise deletion of variables with missing values («= 8) and exclusion of involuntary terminations (n = 13), 234 observations were available for the analysis. Results Table 2 presents the means, standard deviations, and intercorrelations of the variables used in the analysis. As is often the case, the interaction term was highly correlated with at least one of the main effects, in this case, job satisfaction (Darlington, 1990; Gerhart, 1990a). This multicollinearity suggests the inappropriateness of standard regression approaches (discussed later). As was recently pointed out by Huselid and Day (1991), use of ordinary least squares (OLS) regression when analyzing dichotomous dependent variables, such as turnover, often yields inappropriate results. This is because estimates of the effect of independent variables on a dichotomous dependent variable using OLS yield heteroskedastic error terms, violating an assumption of OLS regression (Goldberger, 1991). An often-recommended solution to this problem is logistic regression. However, as is true with respect to OLS regression, logistic regression is very sensitive to multicollinearity (Hosmer & Lemeshow, 1989), which often arises when testing interactions (Darlington, 1990). The presence of multicollinearity often requires corrective procedures, such as ridge regression (Lin & Kmenta, 1982). What was needed for the purposes of my study was a technique that would logistically transform the dependent variable while also dealing with the multicollinearity problem. Fortunately, LISREL 7 (Joreskog & Sorbom, 1989) allows estimation of ridge regressions based on polychoric and polyserial correlations. Polychoric correlations (transformed correlations between two dichotomous or ordinal variables) and polyserial correlations (transformed correlations between a dichotomous or ordinal variable and a continuous variable) allow correction of the distributional properties of dichotomous variables using a log-linear transformation in much the same

4 398 TIMOTHY A. JUDGE Table 2 Means, Standard Deviations, and Intercorrelations of Variables Used in Analysis Variable M SD 1 1. Alternative employment opportunities 2. Education 3. Age 4. Job tenure 5. Wage rate 6. Overall job satisfaction (JS) 7. Affective disposition (AD) 8. JS X AD 9. Voluntary turnover , , *.16*.17*.19* *.15*.10.15*.02.70*.20* * *.05.19* *.92*.54* * Note. N = 234. * p <.05, two-tailed. way that logistic regression transforms dichotomous dependent variables (Bollen, 1989; Olsson, 1979). Thus, this approach meets the same goals as logistic regression, in that it does not waste data (continuous measures of job satisfaction and affective disposition were used), it avoids heteroskedasticity in error terms, and it corrects the distribution of turnover using logistic methods. The advantage of this method over logistic regression is its ability to deal with problems of multicollinearity. The ridge regression results are presented in Table 3. I followed the typical procedure of first entering the control variables into the equation, then entering the main effects, and finally, entering the interaction. In general, the control variables exerted relatively weak effects on turnover, although the direction of the effects followed expectations. With respect to the main effects, job satisfaction significantly negatively predicted turnover whereas affective disposition significantly positively predicted turnover; as a set, the main effects contributed a significant amount of variance to the equation. As hypothesized, the interaction between affective disposition and job sat- Table 3 Ridge Regression Results Predicting Voluntary Turnover Variable entered on step Step 1: Control variables Alternative employment opportunities Education Age Job tenure Wage rate Change in R 2 Step 2: Main effects Overall job satisfaction (JS) Affective disposition (AD) Change in.r 2 Step 3: Interaction JSX AD Change in R *.021 SE ** ** **.185**.053*.076 Note. N = 234. The effect of independent variables on turnover was estimated by applying a log-linear transformation to the turnover variable based on polychoric and polyserial correlations. * p <. 10, one-tailed. ** p <.01, one-tailed. isfaction in predicting turnover was significant. When I entered the interaction in a hierarchical moderated regression analysis, the interaction explained a significant amount of the variance in turnover beyond the effect accounted for by the other variables. The interaction is graphically represented in Figure 1. To facilitate interpretation of the interaction, the methods used to describe the nature of interactions in moderated regression analyses were followed (Cohen & Cohen, 1975; Stone, 1988; Stone & Hollenbeck, 1989). Specifically, the significant interaction indicates that the slope of the regression line representing the effect of job satisfaction on turnover depends on the disposition of the individual. The figure illustrates that for those with a positive disposition, job satisfaction and turnover were significantly negatively related (i.e., the correlation between job satisfaction and turnover for individuals with a positive disposition was -.39, p <.01). Conversely, job satisfaction had only a weak effect on turnover for those with a negative disposition (i.e., the correlation between job satisfaction and turnover for those with a negative disposition was -.05, ns). This also is consistent with the arguments made earlier regarding the nature of the dispositional interaction. Given that past research has suggested that alternative employment opportunities interact with job satisfaction in predicting turnover, the interaction of these two variables was investigated. However, this interaction did not add a significant amount of variance to the regression equation. Because members of this sample were not occupationally heterogeneous, this was not intended to be a formal test of the moderating effect of labor market alternatives on the relationship between job satisfaction and turnover. Rather, it demonstrates that, for this sample, the model was correctly specified in this respect. Finally, because past disposition research suggests that job satisfaction is influenced by affective disposition (Arvey et al, 1989; Staw et al., 1986; Staw & Ross, 1985), a LISREL model was estimated that made job satisfaction endogenous to affective disposition. Whereas the coefficient from affective disposition to job satisfaction was significant, supporting the efficacy of the dispositional perspective, the effects of the other coefficients on turnover were comparable. Thus, the specification reported in Table 3 appears to be valid.

5 DISPOSITION, JOB SATISFACTION, AND TURNOVER Positive Disposition 65.6).25-- Turnover Rate Negative Disposition (M. = 53.2) Job Satisfaction 200 Figure 1. Interaction of affective disposition and job satisfaction in predicting voluntary turnover. (Plotted lines illustrate the effect of job satisfaction on turnover for those scoring 1 standard deviation above the mean on the measure of affective disposition [positive disposition] and for those scoring 1 standard deviation below the mean on the measure of affective disposition [negative disposition].) Discussion This study provides support for Weitz's (1952) hypothesis that affective disposition moderates the relationship between job satisfaction and voluntary turnover. Specifically, employees with a positive disposition who were dissatisfied with their jobs were much more likely to quit than other individuals. Equivalently, job satisfaction and voluntary turnover were more highly related for employees with positive dispositions than for employees with negative dispositions. It should be noted that, although the effect of all covariates was in the expected direction, in most cases the effects did not reach statistical significance. This was somewhat surprising given that past research has suggested their importance. Several limitations in the study may provide explanations. First, because a number of the covariates exerted moderate effects on turnover, but did not reach statistical significance, it is likely that a larger sample size would have provided more supportive evidence regarding the influence of several variables on turnover. Second, several covariates were highly related to each other (see Table 2), which may have diminished the unique influence of each. Third, the sample was not occupationally heterogeneous. It is quite possible that a more diverse sample in terms of wage rates, education, and alternative employment opportunities would have yielded different results for these variables. Finally, Hulin (1991) has argued that the low base rate of turnover makes it difficult for many variables to exert much of an influence. These arguments are well taken, although the distributional corrections applied in my study should have mitigated this concern. Consistent with past research, the results did support the main effect of job satisfaction on turnover, but a positive effect of disposition on turnover also was uncovered in the regression (although it should be noted that the zero-order correlation between affective disposition and turnover was only.05). A potential explanation for this finding may be that those with positive dispositions are more willing to proactively change their lives. Such an explanation is consistent with findings that those experiencing states of positive affect are more likely to proactively change their situation (Isen & Baron, 1991). For example, Isen and Baron argued that affectively positive individuals are more motivated to avoid unpleasant outcomes. This suggests that, for a certain level of job satisfaction or dissatisfaction, individuals in a positive frame of mind may be more willing to take proactive steps, such as quitting their jobs. This reasoning is speculative, however, and future research is required to confirm or disconfirm it. Returning to the hypothesized interaction, there were several conceptual considerations supporting the moderating effect of disposition on the job satisfaction-turnover relationship. It is not possible for this study to demonstrate directly which of these explanations is correct. However, future research should be able to uncover the relative validity of these different predictions. For example, the hypothesis that a dissatisfying job is

6 400 TIMOTHY A. JUDGE more salient, and alarming, for individuals who are generally satisfied than for those who are generally dissatisfied can be investigated by comparing the withdrawal cognitions of those with positive dispositions and low job satisfaction with withdrawal cognitions of those with negative dispositions. Fisher and Locke (1992) suggested that those negatively disposed toward life are less likely to translate job dissatisfaction into withdrawal behaviors than are other individuals because individuals with negative dispositions are not accustomed to acting on the basis of their levels of job dissatisfaction (which may be on par with how they feel about the rest of their lives). Conversely, Fisher and Locke suggested that individuals equally dissatisfied with their jobs, but more positively disposed toward life, may be quite active in changing their work situations because job dissatisfaction is a new and uncharacteristic state for them. Fisher and Locke's results are only preliminary, but they do suggest that it would be useful for future research to address this issue. At first glance, the interaction observed also appears to be consistent with Weitz's (1952) contention that assessment of affective disposition permits a more accurate assessment of relative job dissatisfaction. However, the empirical data are not entirely consistent with respect to this hypothesis. Full corroboration of Weitz's specific prediction requires that individuals who have a positive disposition and are dissatisfied with their jobs have the highest turnover rate and that individuals who have a negative disposition and are satisfied with their jobs have the lowest turnover rate. Only the former of these two conditions was supported, providing equivocal support for Weitz's prediction. It was also suggested that the interaction may be expected because the neutral objects questionnaire reflects a response bias. Although this is an empirical question for future research, past research on the JDI casts some doubt on this interpretation. As documented by Smith et al. (1969), the JDI has not been shown to be affected by response sets. Furthermore, using item response theory, Drasgow and Hulin (1990) indicated that items in the JDI have a very high ability to discriminate between individuals with respect to their true level of job satisfaction. This suggests that most items contained in the JDI are not subject to response sets. Finally, it is not possible for this study to test the prediction that changes in job conditions are constrained by dispositional factors (Staw & Ross, 1985). Although the interaction observed is consistent with this hypothesis, full corroboration would require longitudinal data that directly assess changes in job conditions with corresponding changes in job satisfaction. The results of this study suggest implications for practice and future research. The results indicate that affective disposition is an important construct to consider when one is interested in the prediction of turnover. If these results generalize, the effect of job satisfaction on turnover depends on the propensity of employees to be satisfied in general. To employees unhappy with most things in their lives, job dissatisfaction is not a particularly important factor in decisions to quit, and these employees are less likely to quit when dissatisfied with their jobs. On the other hand, job dissatisfaction is a significant factor in turnover decisions made by employees with positive dispositions. As Weitz (1952) maintained, the importance of job dissatisfaction to turnover depends on the general disposition of the individual. This study provides more evidence regarding the efficacy of the dispositional approach in explaining organizational phenomena. Future research, using a different sample, is needed to replicate the interaction between affective disposition and job satisfaction in predicting turnover. If the results are replicated, it would be useful to investigate if the interaction applies to other behaviors, or to the behaviors cumulatively. For example, Judge and Hulin (1991) found that a number of withdrawal behaviors displayed sufficient covariation to represent a common construct. Finally, research is needed to directly investigate the psychological explanations reviewed above that might account for the effect observed. References Arvey, R. D., Bouchard, T. J., Segal, N. L, & Abraham, L. M. (1989). Job satisfaction: Environmental and genetic components. Journal of Applied Psychology, 74, Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Brief, A. P., Burke, M. I, George, J. M., Robinson, B., & Webster, J. (1988). Should negative affectivity remain an unmeasured variable in the study of job stress? Journal of Applied Psychology, 73, Carsten, J. M., & Spector, P. E. (1987). Unemployment, job satisfaction, and employee turnover: A meta-analytic test of the Muchinsky model. Journal of Applied Psychology, 72, Cohen, J.. & Cohen, P. (1975). Applied multiple regression/correlation analysis for the behavioral sciences. New York: Wiley. Dalton, D. R., & Todor, W D. (1979). Turnover turned over: An expanded and positive perspective. Academy of Management Review, 4, Darlington, R. B. (1990). Regression and linear models. New York: McGraw-Hill. Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, Diener, E. (1990). Challenges in measuring subjective well-being and ill-being. Manuscript submitted for publication. Drasgow, E, & Hulin, C. L. (1990). Item response theory. In M. D. Dunnette & L. M. Hough (Eds), Handbook of industrial and organizational psychology (2nd ed., Vol. 1, pp ). Palo Alto, CA: Consulting Psychologists Press. Fisher, C., & Locke, E. A. (1992). The new look in job satisfaction theory and research. In J. Cranny, P. Smith, & E. F. Stone (Eds), Job satisfaction: How people feel about their jobs and how it affects their performance (pp ). Lexington, MA: Lexington Books. Gerhart, B. (1987). How important are dispositional factors as determinants of job satisfaction? Implications for job design and other personnel programs. Journal of Applied Psychology, 72, Gerhart, B. (1990a). Voluntary turnover and alternative job opportunities. Journal of Applied Psychology, 75, Gerhart, B. (1990b). What is the practical relevance of dispositional effects on job satisfaction? (Working Paper No ). Ithaca, NY: Cornell University, Center for Advanced Human Resource Studies. Goldberger, A. S. (1991). A course in econometrics. Cambridge, MA: Harvard University Press. Hosmer, D. W, & Lemeshow, S. (1989). Applied logistic regression. New York: Wiley. Hulin, C. L. (1991). Withdrawal, persistence, and commitment in organizations. In M. D. Dunnette & L. M. Hough (Eds), Handbook of industrial and organizational psychology (2nd ed., Vol. 2, pp ). Palo Alto, CA: Consulting Psychologists Press.

7 DISPOSITION, JOB SATISFACTION, AND TURNOVER 401 Huselid, M. A., &Day, N. E. (1991). Organizational commitment, job involvement, and turnover: A substantive and methodological analysis. Journal of Applied Psychology, 76, Isen, A. M., & Baron, R. A. (1991). Positive affect asa factor in organizational behavior. In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior (Vol. 13, pp. 1-53). Greenwich, CT: JAI Press. Joreskog, K. G., & Sorbom, D. (1989). LISREL 7: A guide to the program and applications. Chicago: Statistical Package for the Social Sciences. Judge, T. A. (1990). Job satisfaction as a reflection ofdisposition: Investigating the relationship and its effect on employee adaptive behaviors. Unpublished doctoral dissertation, University of Illinois at Urbana- Champaign. Judge, T. A. (1992). The dispositional perspective in human resources research. In G. R. Ferris & K. M. Rowland (Eds.), Research in personnel and human resources management (Vol. 10, pp ). Greenwich, CT: JAI Press. Judge, T. A., & Bretz, R. D. (in press). Report on an alternative measure of affective disposition. Educational and Psychological Measurement. Judge, T. A., & Hulin, C. L. (1991). Job satisfaction and subjective wellbeing as influences on employee withdrawal (Working Paper No ). Ithaca, NY: Cornell University, Center for Advanced Human Resource Studies. Kanter, D. L., & Mirvis, P. H. (1989). The cynical Americans: Living and working in an age of discontent and disillusion. San Francisco: Jossey- Bass. Lin, K., & Kmenta, J. (1982). Ridge regression under alternative loss criteria. Review of Economics and Statistics, 64, Mellow, W (1980). A multinomial logit model of labor turnover. Journal of Economics and Business, 32, Mobley, W H. (1977). Intermediate linkages in the relationship between job satisfaction and employee turnover. Journal of Applied Psychology, 62, Mobley, W H. (1982). Employee turnover: Causes, consequences, and control. Reading, MA: Addison-Wesley. Olsson, U. (1979). Maximum likelihood estimation of the polychoric correlation coefficient. Psychometrika, 44, Parsons, C. K., & Hulin, C. L. (1982). An empirical investigation of item response theory and hierarchical factor analysis in applications to the measurement of job satisfaction. Journal of Applied Psychology, 67, Porter, L. W, & Steers, R. M. (1973). Organizational, work, and personal factors in employee turnover and absenteeism. Psychological Bulletin, 80, Roznowski, M. (1989). Examination of the measurement properties of the Job Descriptive Index with experimental items. Journal of Applied Psychology, 74, Smith, P. C., Kendall, L., & Hulin, C. L. (1969). The measurement of satisfaction in work and retirement. Chicago: Rand McNally. Staw, B. M. (1980). The consequences of turnover. Journal of Occupational Behavior, I, Staw, B. M., Bell, N. E., & Clausen, J. A. (1986). The dispositional approach to job attitudes: A lifetime longitudinal test. Administrative Science Quarterly, 31, Staw, B. M., & Ross, J. (1985). Stability in the midst of change: A dispositional approach to job attitudes. Journal of Applied Psychology, 70, Stone, E.F. (1988). Moderator variables in research: A review and analysis of conceptual and methodological issues. In G. R. Ferris & K. M. Rowland (Eds.), Research in personnel and human resources management (Vol. 6, pp ). Greenwich, CT: JAI Press. Stone, E. F, & Hollenbeck, J. R. (1989). Clarifying some controversial issues surrounding statistical procedures for detecting moderator variables: Empirical evidence and related matters. Journal of Applied Psychology, 74, Watson, D. (1988). The vicissitudes of mood measurement: Effects of vary ing descriptors, time frames, and response formats on measures of positive and negative affect. Journal of Personality and Social Psychology, 55, Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, Watson, D, & Tellegen, A. (1985). Toward a consensual measure of mood. Psychological Bulletin, 98, Weitz, J. (1952). A neglected concept in the study of job satisfaction. Personnel Psychology, 5, Received April 15,1992 Revision received August 24,1992 Accepted August 25,1992